what about hsitorical logs for that node name? Is there anything in puppetserver, puppetdb, or console logs that show that node coming out and going back in?

Nothing that I was able to find. It's clear a PE or puppetserver bug, but how to catch/reproduce it I have no idea

I might need to add something to fail catalog if too few resources

you know what this "feels" like? Sometimes when running a Bolt query consecutively, my node numbers/names ,etc. change. All I have to do is run that command again, and I get the ones it "missed". I wonder if this is similar, like something over pxp isn't getting there before some timeout, and you don't get the value back, but next run you do?

well puppedb here is a bystander, no classes applied, so query is correct

but why it gets an empty catalog - that's a mistery

no data returned?

at least PE agent classes should be included, right?

unless the classifier is the side the timeout happens

well, I do think it's classifier glitch, but opening support ticket will lead nowhere, unless I can reproduce the problem

that's like the mysterious oil leak no one can find. Seems like sometimes there isn't as big/long a run, so it runs fine. Might be when tehre are a lot of changes, so it can't get to everything within a timeout. I'm lookiing for one right now

How long is your Puppet run generally, @vchepkov?

~30 seconds or so

~12 when issue happens

well that wouldn't be it. services default timeout is 30s as is the various http and such timeouts

It would take an engineer to tell us how they do that. I mean, if you've got an open HTTP/HTTPS session and the timeout is reached, do they just slam shut the connect, send a FIN, and say "have a nice day"? Do they allow the session to complete since the 3-way has happened and data is on the move? So many questions.

that's a real "head scratcher". Keep us posted if you find anything

Timeouts are errors and classification errors are fatal to catalog compilation.

yeah, but a node just disappearing from classification until next run is ... well, weird.

yep, no errors, still production environment - no resources

that's it

and I wouldn't expect a "no changes" run to look like that either

Hello there! I have an environment that is managed by Puppet 5.5 and the Puppet CA has expired. Is this the best process to follow: https://www.puppet.com/docs/puppet/5.5/ssl_regenerate_certificates.html#regenerate_ca_and_all_certificates I assume I have no choice but to visit each server that has an agent ? Thanks in advance for any feedback!

if it's already expired, that's likely easiest

yes

if it hadn't expired you'd have been able to extend the CA and update the ca.pem

but once it expires puppet gets all sorts of annoyed about it

yep

time to setup bolt!